a more exacting odds line

[formula_2002]

My original system was based on over 5400 picks, culled from a data based of over 115,000 horses and returned a profit of 7% and a 22% win rate.

Currently, that original system is showing a profit of less then 1%, and a 22% win rate, in 850 picks.

Using the same factors in the original system but a much more exacting odds line, I back-fitted 1,375 races that returns a 33% profit and a 21% win %.

The system shows a profit in all but 3 of the odds ranges of 1 through 20-1.

There have been 17 "live' picks, which started on 3/10/2003.
The live picks show a very large profit, which, of course will get lower.

Various reports for the back-fitted system are being posted on my web page for any one interested.

[andicap]

when you say more exacting odds line, do you mean, making what you used to tab at 2-1 (and thus at 50% overlay, bet at 3-1) -- instead make that horse 3-1 so you need 9-2 to bet him???


If so, that move could help a lot of people using oddslines. Fewer bets, more profits through extra selectivity.

[formula_2002]

Andy, it's much more then that. I now calculate two odds lines.

The first it’s the original odds line, which is based on perhaps 20 or so factors.

I now use a second odds line that uses over 70 factors.

I use the higher of the two odds lines as the "minimum odds".
The two odds line gives two different views of the race.

The average winning odds was 3.64-1, while the new average odds is 5.21-1


Joe M

[anotherdave]

Originally posted by formula_2002

There have been 17 "live' picks, which started on 3/10/2003.
The live picks show a very large profit, which, of course will get lower.



Always interested in this kind of stuff, but 17 picks in 18 days would cause me to go bonkers! Is that just 1 track or several?

AD

[Kentucky Bred]

My experience seems to indicate that once you create a mosaic (new hot word) of sorts from which to draw upon and you can prove at least a breakeven on win bets (which you have done at least to me) the "tweaking" effect of demanding a higher price from your key plays will nearly almost result in a higher ROI. First of all the word "breakeven" in horseplaying means a 15-18% ROI to start. So you have proven a profit curve already. Demanding the horses pay more through dual oddslines should give you a big boost.

Joe, I want to commend you for your dedicated work. I receive your daily emails and have visited your site often. I have seen your constant updates there as well. You have asked for nothing and have stuck your neck out nearly every day. You don't seem to get too up or down with the ebbs and flows of the streaks. Your tenacity is noted.

I bought one of the original Always programs back in the mid 90's. I have respect for the program and those behind it. But as a novice computer person back then, I felt that I was not even coming close to realizing what the program might do. That was frustrating to me. I just sort of gave up out of frustration (and a very poor customer service/training method back then). The $7-$14 downloads per track didn't get me too excited either. I'm glad sharper people than me have continued on the pathway, though. Best of luck with your new adjustments.

Kentucky Bred

[formula_2002]

anotherdave
I don't handicap every day. The 17 picks account for 8% of the races I handicap. That averages out to about 2+ picks for every three race cards.

Thats about what the sample produced.

In the sample, I use 42 different tracks and returned a profit at 30 of them, but I expect those resuts to improve as I accumulate more races.

The system appears to be indifferent to the track played.

I'll be posting some reports based on "race type", Distance, surface and track condition in addition to the track and odds line reports already posted.

[formula_2002]

Kentucky Bred

Thank you very much for your generous thoughts.

I have been doing this for so many years, that as I look back, I find it has become a life’s work.

Although I use the all-ways Bris data files, I write my own programs using their "raw" data factors.


I also find The All-Ways, and Bris people very helpful and responsive.

Thanks again

[Rick]

formula,

I really think that your future results will be much better if you use very few factors. Using 20-70 factors will result in too much noise. Over time I've reduced the number of factors to improve my results, not the opposite as you would expect. But I once believed as you do that using a lot of factors would result in greater accuracy so I don't expect you to believe me. Best of luck in your search for the ultimate handicapping method. I wouldn't try to find one thing that works at every track. I'd separate tracks into at least three groups. But that's just my opinion. I could be wrong.

[formula_2002]

Rick, thanks for the good wishes.

We just wont know until its over.. even then I could be wrong

[Rick]

formula,

What always worries me is that I've seen a number of studies of methods that worked over an awesomely large number of races that still failed on a retest. The probability of such a thing occurring by chance is almost zero UNLESS the whole structure of the problem has changed over time. That's a scary thought, and such situations are not without precedent since they occur in economics all the time.

[formula_2002]

Originally posted by Rick

The probability of such a thing occurring by chance is almost zero UNLESS the whole structure of the problem has changed over time. .

Rick, what does the above mean?

Thanks
Joe M

[Rick]

formula,

What I meant is that we use statistical methods to describe a model of a situation that occured in the past and assign significance levels (probabilities) to the variables. IF the data was chosen from the same sample then we can have some confidence that the probabilities are accurate. BUT if the data is chosen from a different sample at some later time, we can only have confidence in them if we're certain that the relationships haven't changed over time. That would be what's called a stationary process.

In econometrics, this is typically a big problem because many economic variables are non-stationary and the models only work really well over short periods of time. When a really big change occurs (like 9-11) the old rules of the game may not apply at all any more. This is very similar to what occurs in horse racing prediction and it's therefore quite worthwhile to read about some of the problems in that field.

A good example of this would be the change in the value of recency. Models developed 20 years ago would overemphasize this as a factor relative to it's value in today's racing. And we have no guarantee that variable that are underbet today (like early speed) won't be bet correctly in the not-too-distant future. Imagine what would happen if accurate pace ratings were published in DRF and every handicapping book published emphasized their correct usage. Of course this has already happened to some extent, but fortunately most of the information published on pace analysis is emphasizing the wrong things, making unsubstantiated claims, and treating it as some kind of art rather than science. So it's still something we can use for a while longer IF we do our own research rather than rely of the claims of others.

[hurrikane]

Hey, great job Joe.
I'm kind of in the 'school or Rick' on this one. Fewer parameters, less noise, not such a strong backfit.

anyway, good results, hope they hold up for you.

[Foolish Pleasure]

Improve via inclusion not exclusion,
your long term prognosis is healthier.

unfortunately we are approaching an age if we haven't already where any information that is easily objectifiable is virtually irrelevant.

[formula_2002]

Rick, I agree with you 90%

The factors I use are used by many people.
I'm truly impressed at how well the odds correlate to the factors.

But having said that, since many of us have the information, it's extremely difficult to make a profit.

I also agree that it's no too difficult to back fit data to show a profit.

I might be wrong, but I think the analysis of the back fitted results will impact future success.

I think it extremely import that the back fitted data show a profit in as many odds ranges as possible, and I look for statically significance in those odds ranges.

Thanks

Joe M

Ps I'm holding out 5% because I'm never 100% sure of anything.
The other 5% because we have not discussed "risk" , and that impacts some of the above.

[Rick]

formula,

If I understand correctly, you separate your analysis into various odds groups, which is a very different approach than any other I've ever heard of before. Since I've never tried this before, I can't verify that it will lead to any greater success, but even if it doesn't I'll have to give you extra credit for creativity. After all, isn't that what we're really missing (creativity). Almost everyone repeats the same old thing over and over. I've discovered that a lot of things work pretty well once you have the right contrarian frame of mind that rejects the obvious. It's not really all that hard. All you have to do is realize that the average person wants a sure thing verified by several different methods before he will bet. So, it's actually more important to know what your opponents are doing that is wrong than it is to know what you're doing is right. I know many will disagree with me but that's to be expected.

[sjk]

Rick,

I do not believe that the nature of the game has changed (although the definition of recency has probably expanded). I developed my program from data from one circuit in the early to mid 90's and it is gratifying that it seems to work as well 5-10 years later at any the track I play. What we are all trying to evaluate is the quality of effort of the animals running and the way the animals run does not change that quickly over a period of a few years.

[Rick]

sjk,

Almost every profitable spot play that I played 20 years ago involved horses coming back in 5-7 days. Now, some of them still work with looser requirements for days away. But, things have definitely changed and I think it would be irresponsible to not recognize it and warn the newer players.

But maybe you're right in saying that things haven't changed that much in the last 10 years. Actually, I'm pretty amazed that early speed works at all after this long. I would have guessed that we would need to play something opposite to this by now. But things last however long they last and we'll just have to react to the real situation. I'm ready to change whenever the public does. Are you?

[formula_2002]

Rick
I got the method of statically testing odds ranges from a book written by B. Fabricand. His book "Horse Sense" is an excellent case study for this method of analysis.

Somewhere around my house I have two copies of the book.

I'm sure it was published in the early 1970's or perhaps late 60's.

I'm also most certain he is an economist

[sjk]

I only play what I perceive to be overlays. So far I am able to play more races than not. As time goes by, I would not be suprised to find fewer overlays to play. I will enjoy it as long as I can.

[Foolish Pleasure]

Rick,
your assuming recency is less potent because bettors are getting smarter.

it could be the other side of the equation, given the soft US economy, owners/trainers could be more desparate. So instead of knowing you got a live animal back off a rush job, your getting a dead animal looking for minor awards to enable their connections to just get by.

Think things aren't tough? Note the quality claiming trainers dropping horses the last few days each month of the past 6 months or so. Clear sign, come bill time they are cash poor.

[Rick]

FP,

No, I don't assume in the case of recency that the public is smarter, just that the impact values for smaller days away have decreased over time. Actually, it would be easier to argue that they're dumber because they're now underbetting the horses that are off 31-45 days. I've heard people mention a lot of reasons for the change including drugs and better training techniques, but the reason really doesn't matter. If the value of a factor changes over time, then we'd better stay up to date.


formula,

I'm familiar with that book and others by Fabricand. An interesting guy even if I can't figure out what the rules really are for the "principle of maximum confusion". At least it was named correctly.